home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Power Tools 1993 November - Disc 2
/
Power Tools Plus (Disc 2 of 2)(November 1993)(HP).iso
/
hotlines
/
gsyhl
/
sunbg.txt
< prev
next >
Wrap
Text File
|
1993-09-15
|
62KB
|
1,137 lines
H P C O N F I D E N T I A L
= = = = = = = = = = = = = = = = = = = = = = =
Sun SPARCserver 1000 and SPARCcenter 2000
C O M P E T I T I V E K I L L E R K I T.
Written by Simon Pearson
EMC Sales Centre, Boeblingen.
Contributions by Todd Thiemann
GSY Delivery Marketing, Cupertino.
C O N T E N T S
===============
1. What is Sun saying?
2. HP's Counter Arguments.
3. The Situation.
- Moving the Goalposts
- Symmetrical Multi-Processing
4. HP on the Offensive.
- Systems Management
- Systems Availability
- Limited Product Line & Poor Upgrades
- Close RDBMS Vendor Relationships
- Customer Support
- Proprietary, poor quality Solaris
- Sun's Financials
- SPARC's Future
5. Product Positioning
6. Consultants Quotes Page
1. What is Sun saying?
=========================
Sun is making some very aggressive attacks on HP with their
SPARCserver 1000 and SPARCcenter 2000 commercial computing
systems. Here's what they're saying with some ready
counter-arguments. We'll look closely at the substantiation of
these arguments in the remainder of this paper.
A: "Ss 1000 and Sc 2000 SMP performance is better than HP's
Series 800 and is more scalable too".
CA: "HP's highest TPC-A figure is 711, Sun's is only 401, and Sun
achieved that by using the maximum no. of available cpu's (8)
and a beta version of the operating system that is not yet
available. Sun only offer mid-range boxes and have nothing
to offer high-end customers".
A: "Easier and cheaper upgrades, future upgrades o faster cpu
chips simple".
CA: "Sun 2-cpu card upgrades offer a narrow upgrade path dictated
by Sun Microsystems. If you want more cpus, you are forced to
pay for memory and I/O slots. If you want I/O, you must add
another CPU board. And remember that 2 CPUs provide negligible
power. You need at least 4 CPUs to make an impact. HP provides
better granularity by giving the customer exactly what they
want."
A: "Outstanding I/O Pipeline with XDBUS and Dual XDBUS - fast
systems with room for growth".
CA: "HP's bus is 4 times faster than Sun's, and Dual XDBUS is a
long way from being deliverable".
A: "Sun has really close RDBMS vendor relationships (ORACLE) -
much closer than HP's".
CA: "HP is now Oracle's largest hardware platform and has been
the UNIX sales leader for the last 5 years".
A: "Sun has its own internal experiences/references in using
SPARCserver and SPARCenter systems".
CA: "Yes, but do they have anyone else? Ask Sun to give you
serious customer references who have been using SPARCservers
and SPARCcenters in comparable environments for over a year."
A: "Sun's SMP offers higher availability, reliability, uptime,
and fault tolerance than HP's Series 800"
CA: "Sun's high availability offering is limited and is not as
robust as HP's. The more CPU's you have, the higher the
chance of failure. HP offers full high availability, better
uptime, and better reliability than Sun's systems".
HP ON THE OFFENSIVE.
====================
Use these arguments in conversation with customers, and force the
Sun Sales Rep to answer them.....
A: "Sun's SPARCserver & SPARCcenter is a VERY limited product
line that only covers the mid-range. Sun have nothing to
pitch against HP's F or G Series Novas or Emerald-class
systems, and don't even have compatible low to high-end boxes
in their range."
(See: Limited Product Line & Poor Upgrades).
A: "Sun offers poor investment protection. You cannot use
previous Sun disks and memory with the Ss1000 and Sc2000, and
you cannot use Ss1000 internal disks or memory with the Sc2000."
Sc2000."
(See: The Situation)
A: "The SPARCcenter 2000 offers few advantages over the
SPARCserver 1000, that's why Sun isn't pushing it."
(See: The Situation).
A: "Sun's Solaris is the most proprietary version of UNIX
available today".
(See: Proprietary, Poor Quality Solaris).
A: "Sun has a reputation for bad software quality. Solaris 2.0
had more than 1,000 bugs in it, and current Solaris is not of
commercial quality (although it makes for good benchmarks)."
(See: Proprietary, Poor Quality Solaris).
A: "Sun cannot support you properly, they are downsizing
personnel and have been quoted as viewing support as a time
consuming costly activity that diverts attention from getting
order dollars. HP is renowned for its quality worldwide
Support".
(See: Customer Support).
A: "Sun is in financial trouble, profits are down 10%, their
Chief Financial Officer admits they're only selling upgrades,
and the market has not accepted their new systems".
(See: Sun's Poor Finances).
A: "Most industry analysts are saying SPARC has already reached
the end of its lifecycle".
(See: SPARC's Future).
A: "Sun can't get Solaris to scale beyond 8 cpu's, and can't get
XDBUS to work at all".
(See: Symmetrical Multi-Processing).
A: "Sun offers only a mid-range system with a poor bus, limited
scalability, a buggy operating system, a lack of commercial
software and functionality, poor investment protection, poor
support, poor quality, and no references....does that sound
like a good deal?"
THE SITUATION.
==============
Let's look at the situation realistically:- HP is No.1 in UNIX,
and being No.1 makes you everyone else's target. HP is the
standard that everyone else is strives to achieve, HP has the UNIX
revenues everyone else would like to have, and HP is in the
industry position to which all our competitors aspire.
HP is No.1 because we made the right bets a long time ago - heavy
investment and focus on the UNIX market, the development of
PA-RISC technology, and the early availability of applications.
HP was in early and gained a strong foothold in the burgeoning
UNIX market, and it is natural that now HP has become the No.1
target for the increasingly aggressive competition.
This paper focuses on one such competitor, Sun Microsystems, and
will equip you with the necessary information and knowledge to
effectively compete and win against Sun, stopping them in their
tracks. Given HP's advantages over Sun, we can and should win
against Sun.
Why is it that Sun is trying to break into the commercial UNIX
server market? This is the key question that will help you to
understand where Sun is coming from and what it has to offer.
Sun is a workstation vendor. They grew up with workstations and
initially gained the lion's share of the market; but recently they
have been squeezed at the high-end by HP and Silicon
Graphics, have lost market share, and their attempted push of the
market to low-end commercial workstations has been blocked by
Intel and Microsoft's NT. With a decreasing workstation market
share and a fiercely competitive outlook, Sun has been forced to
search for new market opportunities, and in fact they first tried
penetrate the commercial server market two years ago with their
SunServer and 600MP boxes, achieving little success and only
selling these systems as file servers to the Sun workstation
installed base. With poor finances, declining margins and
decreasing market share, Sun is in panic mode and will do anything
to win sales - any price, any conditions. We've recently seen Sun
going to extreme lengths to win business - huge discounts of up
to 80%, giving away standby systems for CPU failover, offering
free-of-charge upgrades to future generation SPARC CPU chips, and
even a case where they offered to buy the customer an HP Emerald
or any other competitors' box and give it to them free-of-charge
if Sun's solution didn't work out for them! Some of this sounds
pretty tough to beat, but it also sounds pretty desperate, doesn't
it?
You should point out to your customer that Sun is fighting a
rearguard action. Their final recognition of Motif and DCE as the
standards of the future clearly demonstrates the failure of their
OpenLook and ONC, they are now trying to play with the big boys in
commercial systems computing, yet have no experience in the
market, as pointed out in a December 1992 Mainframe Downsizing
Paper from Workgroup Technologies Inc, that positions Sun among
a group of pretenders:- "Sun has no experience in the data center
and niche vendors such as Pyramid, Sequent, and Data General with
their limited 3rd party software offering, weak data center
management tools, and poor corporate financials can only be
considered for point, non-strategic downsizing opportunities with
low risk".
Sun is trying to extend their SPARC workstation technology into
the mid-range commercial systems market. They are trying to get
everyone to believe that the SPARCcenter 2000 is a Data Center
system, which it is not. The systems' performance, bus speed,
lack of commercial software, lack of commercial tools, and limited
expandability show that it is not a Data Center oriented system
at all, and demonstrate that as a workstation vendor Sun is simply
trying to 'talk up' their boxes into the mid-range. They are
initially focusing their commercial offering on their workstation
installed base to try to gain installed references, but seem to be
achieveing limited success. HP understands what the Data Center
is all about; Emerald was designed from the very start to deliver
thousands of TPS, a mainframe-speed bus, excellent commercial
software, rich systems management tools, and limitless
expandability. Emerald is THE NEW GENERATION 3090 MAINFRAME WITH
NONE OF THE DISADVANTAGES.
Extended workstations are not part of this market and simply
cannot compete, but as a workstation vendor Sun will not
understand this. The Solaris operating system demonstrates their
lack of understanding of what it takes to compete in the
commercial world - it offers no Data Center functionality and a
level of quality/robustness that's simply unacceptable for
commercial users.
Sun's previous success has resulted in large part to achieving
large volumes that in turn drive down manufacturing costs. What
few customers recognize is that the SPARCcenter 2000 is a low-
volume product that Sun uses to demonstrate a broad product line.
As a low volume product, Sun will devote few resources to
expanding the SPARCcenter 2000's capabilities. Sun sold 58,000
workstations in the 4th quarter of fiscal 1993, but only 500 Sc
2000s. Where do you think Sun's development effort is focused?
In addition, the SPARCcenter 2000 seems to be almost a dead
product. Sun confuses the SPARCserver 1000 and SPARCcenter 2000
positioning and seldom pushes the SPARCcenter 2000 anymore.
The Sc2000 has a slow dual bus that Sun can't get to work,
doesn't scale beyond 8 CPU's, and coming strongly from a
workstation background was the wrong architecture to start with
anyway. Sun HAD to introduce the SPARCserver 1000 to save face and
keep moving, has stopped seriously pushing the SPARCcenter 2000.
Tactics such as these may work well in the workstation market but
with such a short product life cycle Sun has left commercial
customers hanging with the SPARCcenter 2000. Sun's Sales Force
is pushing the SPARCserver 1000 strongly, and the only
significant thing they can bring to the table is price. The list
prices of SPARCcenter 1000 and HP's Nova family are in fact very
close, but Sun's tactic will be to load the configuration with RAM
and disk - areas where HP is not as competitive, in an attempt to
price HP out of the deal. Do not let Sun have your customer break
everything down to the component level, focus on the overall price
of the total solution and don't get sidetracked into competing on
a RAM card and disk spindle level. It's the total price that
counts. Sun will be forced to bid big discounts to compete with
us and we should reply with equal aggressiveness.
Moving the Goalposts.
_____________________
Having lost the arguments on technology and price, Sun will try to
move the sale to things that they simply can't win with today.
Coming from a workstation background Sun is comfortable talking
about futures: future CPU's, future performance estimates, future
disks and RAM, and future field-upgrade pricing. The good news is
that HP has a better product evolution than Sun can even dream
about, and what's more our claims are backed up by a track record
of being able to deliver - Sun's are not.
Today HP's Nova family in 2-way CPU configuration delivers more
performance than Sun's high-end 8-way system (411 TPC-A versus
400.8 TPC-A respectively), and in November '93 the story will get
even better with 8-way Emerald TNT. January 1994 will show 25-30%
performance improvements in the Nova line that will leave Sun in
the dust.
Symmetrical Multi-Processing.
_____________________________
The Sun Sales Rep's next line of attack will be their Sun's SMP
story, however Sun does not believe in SMP - they have simply been
forced to implement it because of the poor performance and high
complexity of their SPARC CPU architecture. And now Sun is trying
to make a virtue out of necessity by claiming MP is the wave of
future. Performance of the SPARC chipset is lousy, way behind
IBM's Power architecture, DEC's Alpha, and HP's PA-RISC. Losing
to HP and Silicon Graphics in the workstation performance fight,
Sun has been forced to implement SMP to cover up its poor chip
design, and this is a situation Sun would clearly prefer to avoid.
The manufacturing cost of an 8-way system is significantly higher
than that of a uniprocessor or dual-processor system, and higher
profit margins narrow Sun's profits(Sun's latest financial results
show nearly a 30% increase in cost-of-sales). Contrast this with
balanced systems offering:- the fastest uniprocessors in the
industry and SMP when you need it, but without the disadvantages
inherent in Sun's 'patched-together' offering.
As a workstation vendor Sun don't understand symmetrical
multiprocessing, and have a VERY poor reputation for delivering
SMP scalability. They seem to be repeating their past mistakes;
witness the poor scalability figures of the SPARCserver
1000/8-way's 400 TPC-A (50Mhz chip) against the Ss 10's 108 TPC-A
with 1 40Mhz cpu. In discussing Sun's poor SMP scalability, DH
Brown Associates state:- "performance will be embarrassingly weak
- as Sun found out when it took exactly this approach with Solaris
1". Previous "four processor configurations frequently achieved
only poor performance - almost NEVER approaching the 'linear
performance increase' hoped for, and sometimes even DECREASING
total throughput compared to the two processor configurations".
DH Brown continue by saying they are sceptical Sun will achieve
good SPARCcenter performance in the near term as vendors such as
DG, NCR, and SEQUENT with long multiprocessing histories and
experience have taken YEARS to tune their SMP UNIX implementations
to get near-linear performance increases. They go on to say that
anyone believing that Sun can immediately achieve what's taken
other vendors with much greater experience years to achieve is
naive at the very least!
Point out to your customer that Sun's 400.8 TPC-A rating for the
SPARCserver 1000 was achieved using a Beta version of the
Solaris operating system not yet available to customers - Solaris
2.3, and that today's Solaris 2.2 is estimated by Gartner
Group to perform up to 30% slower; that although Sun claim Solaris
2.3 availability sometime late 1993, industry-watchers believe it
will be 1st half 1994 before it ships. Sun have been known to use
Solaris 2.3 against HP-UX 9.0 in real-life customer benchmarks -
thereby pitting a future, not-yet-stable, and potentially buggy
operating system against UX 9.0 that IS stable and has been
shipping for 10 months now into real customer installations. Sell
HP's proven, audited, and deliverable TPC-A figures hard and don't
let Sun get away with trying to convince your customers that they
are today's performance leader.
However, Sun is being pretty aggressive with their benchmarks;
where they have beaten HP is in database applications that are
compute-intensive with little or no OLTP or batch jobs (just what
you would expect from a compute-focused workstation vendor), and
where Sun have had an early start and have been able to tune the
software to run better with SMP. We're finding however that there
are very few commercial customers whose profiles match even
loosely Sun's strengths:- most large database users need overnight
batch updates/downdates and use OLTP extensively in their
day-to-day processing. Typical systems workloads are one-third
batch and two-thirds OLTP, and this is good news for HP as with
faster processors our systems have much higher commercial OLTP
capabilities and batch performance, which typically uses a single
cpu. When discussing this with your customer focus on the TPC-A
per cpu argument that the Sun Sales Rep has been glossing over,
get your customer to demand a benchmark that includes at least
one-third batch, and help them to understand that Sun's boxes
simply can't handle the commercial workload expected of them:-
Sun Ss1000/Ss2000 8-way 50 TPS per processor
HP Nova Model 60 1 way 280 TPS per processor
Sun's audited benchmark results highlight some interesting
scalability issues they're facing. The 40Mhz chip in 1-way cpu
configuration produced 108 TPC-A in the Ss 10 Model 41; yet the
50Mhz chip in 8-way guise produces only 400.8 TPC-A. Does this
demonstrate good scalability? Is Solaris the limiting factor? Is
the XDBUS really ready? - Is it shipping even? ...and what
happens when your customer needs more than 400.8 TPC-A? What will
Sun offer and when? Can they prove it? And why don't they have
it today? Show your customer that HP's Emerald systems produce an
audited 711 TPC-A today, that HP's Model 70 Nova's deliver 411
TPC-A with only 2 cpu's, and that Sun's 8-way SPARCserver at 400
TPS is a weak offering by comparison.
HP on the Offensive.
____________________
Having stopped Sun in their tracks by demonstrating HP's
technology leadership, aggressively countering their price, and
proving HP's performance advantages, we can then differentiate
with things that Sun can't deliver, or even make claims to today.
Point to the huge portfolio of HP customers running large
commercial UNIX systems on proven HP PA-RISC technology and HP-UX
operating systems. Point to HP's more than 20 years experience in
commercial systems computing, HP's largest market share in
mid-range and high-end commerical UNIX systems, and the references
we have running mission-critical computer installations on HP's
systems. Then ask your customer to get Sun to do the same.
You'll find you can create a feeling of comfort, security and
well-being with HP, and leave your customer feeling exposed if
considering Sun.
Point to HP's sheer breadth of applications software - commercial
applications software - that runs on HP-UX. Ask why it is that
companies like Computer Associates, SAP, Dun & Bradstreet, and
Software AG have chosen HP as THE platform to release UNIX
versions of their mainframe software on? These established
companies need a stable, future-proof platform with the largest
market share to launch their software on, and clearly HP offers
exactly that.
Because Sun has no installed base in the commercial market, and
because there is little demand for their systems (Gartner Group:
"Sun's past commercial conquests focused on point solutions. Its
systems are not often deployed as enterprise systems and Sun does
not appear organized to sell to enterprise-level management...")
they will be able to commit to very fast delivery, typically
within one week; and in order to gain market share and references
they will 'buy the business' by being ultra aggressive with
discounts and give-aways. But fast delivery and a cheap price
often don't outweigh the risks involved in buying a Sun solution,
this is especially true the larger the system is and your
customers should feel uncertain of what they're getting when
something is so cheap. Force Sun to take your customer to any
reference who's been using a Sun commercial server to run their
business for even one year - they can't, and Sun simply aren't
able to deliver what it takes to make a buyer feel safe.
We'll now look in detail at some areas where you can attack Sun
and capture the winning advantage:-
i) Systems Management.
At HP we know how hard it is to sell large, high-end UNIX systems
into Data-Center environments. Typically you're talking to
sophisticated users who are used to rich, functional operating
environments with all the systems management tools and utilities
you can dream of. Some are coming from mainframe environments
where teams of well-resourced Systems and Network Managers keep
multiple mainframes running smoothly using tools like CA's
Unicenter, or IBM's CICS. It's a fact that the larger the system
is, the more concerned the customer will be about the ease of
systems management and getting maximum uptime from the systems.
Sun have a ridiculously weak offering in this area - DH Brown
Associates state that "Sun historically provided the WORST system
management in the UNIX world".
This is an area where HP REALLY wins through and can set the
agenda against Sun, forcing them onto the defensive. HP-UX is
without doubt the world's most developed and functionally rich
UNIX operating system; it's not only richer in features than Sun's
Solaris, it's also more stable and bug-free. In short, Sun's
Solaris simply cannot match HP's commercial UNIX offering in
almost any area. We won't look in detail at each area here, but
if your customer's considering a large commercial system from Sun
you should ask him how he's going to monitor and tune the
performance of his system, how he's going to do any capacity
planning or trend analysis, and what plans he has for storage
management, cpu failure, and job and print management. You can
sell HP's full offerings hard in this area, many computer-industry
people overlook just how much of a headache systems management is,
Sun certainly have - customer's don't. The very best Sun has to
offer is, in many cases, unintegrated 3rd party solutions or their
own complex Solaris operating system routines.
Sell OpenSpool for print job management, and TaskBroker for
networked systems job management. Want to know what Sun offers?
- Traditional, complex UNIX spooling commands and routines with
some 3rd party capabilities if necessary.
Any large system is likely to have large databases and big volumes
of data - sell OmniBack and OmniBack Turbo with LVM along with
HP's IBM 3480 tape drive support, or CA's Unicenter product.
Again, all Sun can offer are traditional, complex UNIX utilities
or a Sun product named CoPilot with online backup for up to 9.3Gb.
Compare that to OmniBack Turbo's 25Gb per hour and Sun starts to
look pretty weak. An HP user with a 50Gb database can back it up
in close to 2 hours - Sun requires over 6 hours! Again, DH Brown
Associates state "Solaris 2 sees great weaknesses in Security and
Storage Management. In Security, Solaris 2 offers only C1/UNIX
security - significantly lower than the C2 and B1 levels offered
by most competitors. Storage Management is an even bigger
problem. Essentially NO Storage Management tools such as disk
managers or enhanced backup/recovery systems are shipping for
Solaris - compared to the army of such crucial tools available for
both Solaris 1 and all competitive UNIX's".
Sun don't have LVM (basic functionality bundled free-of-charge
with HP-UX 9.0) but instead offer Online DiskSuite for an
additional $2,500. It does similar things to LVM, but in a recent
report IDC stated it to be "not as elegant, or as flexible as
LVM".
Another area to hit hard is System accounting and configuration -
how many customers have told you what a great product SAM is? Sun
only offers traditional UNIX systems management utilities that are
neither flexible or easy to use increasing the systems management
headache for their users.
ii) System Availability.
System availability and fault management are KEY topics for not
only large-scale systems users, but all systems users. Sun are
being super-aggressive positioning their SMP systems as almost
fault-tolerant...they're not. Their story sounds good: when one
cpu fails the system can reconfigure itself around the failed
component and restart; but it's not as glossy as the Sun Sales Rep
would have your customer believe. The system does go down, does
need to reboot itself, all the users need to logon again, and
finally the system needs to be taken down again when the engineer
comes to fix the failed component. This is high availability at
its most basic level, it's nowhere near as as robust an offering
as HP's SwitchOver/UX. You should sell HP's full and robust
systems availability and redundancy offerings like Switchover and
RAID hard against Sun in this area.
What about cpu reliability? If you assume a single cpu has a
probability of failing once a year, with eight cpu's you're eight
times more likely to suffer cpu failure - is that why Sun had to
make sure the system could reconfigure itself around failed
components? If your system is going down that often you've got a
major headache. Force a downtime calculation and get your
customer to relate this to their business in terms of lost
business potential/customer dissatisfaction/additional cost; and
then relate HP's outstanding reputation for quality, low
likelihood of cpu failure and fuller high availability offering;
and any customer, not just those running mission-critical
applications will see that HP is serious about high availability.
It's not just a casual gimmick given to you by the number of cpu's
in the box. It's also worth pointing out to your customer that HP
has ranked 1st in the Datapro survey for the last 9 years for
Customer Satisfaction - Sun has managed only 4th.
Bring PowerFail AutoRecovery, MirrorDisk/UX, SwitchOver/UX, RAID
disks, and even Predictive Support into any discussion on high
availability. In addition, you can raise Sun's poor reputation
for hardware and software quality as an issue, we'll talk more
about this later, but in summary Systems and Network Management is
a KEY DIFFERENTIATOR for HP against Sun. HP provides commercial
UNIX systems management functionality and reliability in a way Sun
simply cannot match. Quote Gartner Groups' statement that HP-UX
is "the most robust UNIX environment available" and DH Brown
Associates' praise of HP for the 'high level of quality inherent
in HP-UX".
DH Brown summarized by saying: "Solaris 2's current weaknesses in
multiprocessing implementation, Storage Management, Security, and
other areas illustrate the company's macro challenge - its ongoing
transition from a technically focused workstation company into a
general systems company that addresses both technical and
commercial markets. While begun, clearly much remains to be
accomplished". Sell HP's commercial systems experience hard, do
not let a price-aggressive upstart like Sun soft-sell your
customer into believing they are a commercial player today - they
have a lot of hard work and catching-up to do, it will be painful
for them AND for any customer price-sensitive enough to actually
buy one of their boxes. A lot will be missing, mostly everything
that makes a commercial systems installation successful - things
that HP has, understands, and has proven time and time again with
every Series 800 that's ever been sold. I have a copy of Sun's
own internal competitive slide presentation positioning the
SPARCcenter 2000 against HP. They are honest enough to admit they
have much lower disk expandability, no RAID capability, a weaker
multi-user story, and very poor uni-processor performance!
iii) Limited Product Line and Poor Upgrade Path.
With the SPARCserver and SPARCcenter Sun only offers a product
breadth of 8 times the performance, and 4 times the I/O expansion,
while HP's Series 800 offers more than 15 times the performance
and 60 times the I/O expansion! Sun have nothing to bid against
HP either at the low end or at the high end. HP's F and G series
Nova's offer a lower entry point, and a lower price point than the
SPARCserver 1000, and at an audited 711 TPC-A Sun simply have
nothing to compete with Emerald.
Sun however, will try to position the SPARCserver and SPARCcenter
against our Emerald systems, promising the claimed future
performance estimates - but none of this is deliverable today, and
remains simply a claim. By the time Sun do manage to squeeze more
performance from SPARC, HP will still be in the lead with the
1,400 TPS 8-way PA7100 Emerald, leaving Sun even further behind in
the commercial performance race. However, some customers will
listen to what Sun are saying and will try to get you to bid-up to
an Emerald against Sun - DON'T DO THIS since you will
unnecessarily price yourself out of the bid. A SPARCserver/center
8-way is not comparable in any way to an Emerald system - in
performance, expansion capability, software availability, tools
availability, reliability, or support offering. You should only
bid the Nova range against Sun's products, Emerald is a
Data Center class machine that Sun cannot presently match.
Sun is offering up to 8-way SMP today, and claiming up to 20-way
SMP in the future. They're making aggressive claims for their
SPARC chip technology and are positioning themselves as THE
technology leader with SMP. Focus your customer on WHAT Sun can
deliver today: 8-way SMP; and ask WHY Sun has taken the SMP route
for mid-range UNIX servers as well as at the high end. The
reason, quite simply is that SPARC is WAY behind the performance
curve, and cannot match what PA-RISC and other vendors deliver
today. The SPARC chips' 50Mhz speed is incredibly slow by
comparison with most other RISC implementations, and SPARC's cpu
performance simply can't match technologies like PA-RISC. Many
industry experts believe that SPARC is at the end of its
life-cycle, and that because it is such a complex RISC design
50Mhz speed is the fastest it will ever get. Sun therefore had no
option but to find other ways of delivering acceptable
performance, hence SMP. Even Sun's own internal document admits
SuperSPARC is way behind PA7100 in CPU performance. Any customer
considering a Sun system should be incredibly worried about the
future of their purchase - where will SPARC go, how can I upgrade
my system, who is going to design and manufacture the chip, and
when will it be deliverable? Get your customer to question Sun as
to WHEN Sun will be able to deliver higher CPU chip performance,
WHAT it will be, and HOW MUCH it will cost.
HP's Nova family offers the best upgrade granularity of any UNIX
mid-range system available today. Customers can easily upgrade
performance, expansion or both in a simple, cost-effective manner.
Customers can upgrade performance with simple card-swaps, an
upgrade from the lowest to the highest performance box - a G30 to
G70 card swap for example, is a trivial exercise; and adding I/O
expansion is equally simple through field upgrades through the
F-G-H-I range. Explain the ease and simplicity of HP's upgrade
path to your customer, and show them that HP offers exactly what
they need.
Compare Sun's upgrade path to HP's. A Sun user moving from a
previous Sun machine to the SS1000 must lose all their memory and
most of their disk investment while HP customers maintain theirs.
A Sun customer moving from the SS1000 to the SC2000 loses their
internal disk and CPUs, while the comparable move to a 70-class
machine maintains disk investment.
Sun bundles product into their upgrades that customers don't
always need, but have to pay for. Sun will claim they offer
inexpensive upgrades at $10K per CPU, but the minimum you can buy
is 2 CPUs and these come with RAM bundled at a minimum of
$28k. Since adding 2 slow CPU's to your box is unlikely to do
much to improve performance, realistically most customers will
have to buy at least 4 extra CPU's at a cost of $56k, thereby
pushing Sun's upgrade costs even higher - much higher than HP's
for a similar performance or I/O upgrade.
iv) Close RDBMS Vendor Relationships.
Let's remind ourselves and our customers of some relevant facts:-
In Oracle's FY93 (May 31st, 1993) Oracle sold more licenses on
HP-UX than on any other hardware platform. HP-UX overtook DEC's
VMS to become Oracle's No.1 hardware platform, AND HP-UX has been
Oracle's UNIX sales leader for the past 4 years. Oracle's license
revenues grew 38%, and Oracle's HP-UX revenues grew 58%. With
HP-UX revenues growing twice as fast as all Oracle's other product
lines combined, HP is taking market share from our competitors.
Today Oracle's combined HP platform revenues account for over 14%
of Oracle's FY93 license revenues.
Sun will push the Ss1000 and Sc2000 hard when Oracle are involved
in the deal, but we should push back even harder. Sun will point
out that Oracle is developed on Sun workstations so it must be
naturally suited to Sun's systems - it IS, and it's NOT! Oracle
do develop on Sun workstations, because Oracle's CASE tools
originally ran on Sun; but these workstations run SunOS not
Solaris, and there are considerable differences in the commands in
each operating system making them totally incompatible. Porting
from SunOS to other versions of UNIX like HP-UX is in fact
sometimes easier than porting to Solaris, as a June 1993
Computerworld article reporting on SunOS to Solaris migration
stated:- "large user sites have found that it takes weeks to port
and debug applications. Others anticipate an effort of several
months".
v) Customer Support.
Sun is so bad at customer support their customers made them vow to
make improvements! This was only last year and since then Sun has
been downsizing personnel both in the USA and in Europe - this
doesn't seem to back up their vows at all.
Sun's hardware maintenance is done by 3rd parties, not by Sun, and
although Sun has announced some pretty aggressive levels of
support for customers buying the SPARCcenter 2000 (24 hours/day, 7
days/week, with a 2 hour response time, and 15 days of consulting
free-of-charge upon installation), this is not available in every
country and some customers may have to pay a lot extra to get this
level of support. Confusingly, it is not available on the
SPARCserver 1000 that Sun's pushing so aggressively today.
Gartner Group's Sept. 1992 report states:- "Sun views sales
support, systems integration, and user needs assessment as 3rd
party, time consuming, costly activities that divert from pressing
needs to capture volume orders".
vi) Proprietary Solaris and Poor Quality.
Sun's past close links with Unix International and AT&T gave them
an image of openness and technological leadership, and indeed some
products such as Sun's NFS have been an unqualified success. But
Sun only adopts standards when they are based on technologies
developed at Sun, they will try to push these as de facto
standards and will ignore, as best they can, everything else. Sun
has become increasingly isolated as far as open systems standards
and software is concerned; whilst all other vendors actively
support key industry standards such as OSI, OSF, NCS, Motif, DCE
and DME; Sun is still promoting their own OpenLook and ONC
software, and are yet to endorse DME. In a recent report, DH Brown
Associates said "most of ONC's services are technically weak in
relation to those of DCE. They are much less functional, less
efficient, less secure, harder to use, and harder to program.
Though Sun has begun to move beyond ONC with ONC+, DCE rather than
ONC+ is THE foundation for the next generation of high-function,
heterogeneous distributed computing". Sun's UNIX windowing
software, NeWS, was universally rejected by the market in favour
of X.Windows.
It has often been said that Sun's Solaris is the world's most
proprietary version of UNIX, and if true, this largely cancels out
many of the benefits that open systems have to offer, since a
customer will be locked in with Solaris almost as surely as with a
proprietary operating system.
Contrast this strategy with HP's - HP really does actively
participate in standards organisations all around the world, AND
adopts industry standards as they are developed. In fact, HP has
developed some major parts of the technologies that have in turn
become industry standards - DCE, DME, and Motif. This
demonstrates HP's real-life commitment to Open Systems standards,
and benefits your customer by giving them the best, widely
available tools and software with NO proprietary lock-in.
Solaris quality is a very weak area for Sun when compared to HP,
and should be brought into any discussion on high availability and
downtime as a matter of course. Solaris has not always been
backwards-compatible with previous versions, making users
recompile applications and fix bugs as they arose. As Sun's
primary focus has until now been workstations, operating system
quality and reliability has never been a priority for them, and
Solaris quality today is still unacceptable for commercial users.
Sun admitted that Solaris 2.0 had over 1,000 bugs in the operating
system and they had to offer a "Bug-Fix" subscription hotline
(which they charged users for) - does this sound like the way
things get done in the commercial world? Does it demonstrate
quality software and support? Sun is well known for premature
Operating System releases, and although Solaris 2.3 (due out soon)
is their 3rd version of SMP it's anticipated that it will not
provide the quality commercial customers are expect. Questionable
operating system stability has to be a key issue for anyone unwise
enough to be considering a Sun solution - question your customer
on how much downtime they can put up with, and more importantly
how much data corruption they can afford to suffer. Two day's of
data corruption in your customer's business may wipe out any
difference in price between the Sun and HP systems.
In talking about user migration to Solaris, Michael Zadig,
Director of Sales & Development Engineering at SunSoft (which
designs and sells Solaris) said "The message has been communicated
very clearly, but sometimes there is a latency in the marketplace
in moving to a new technology". Want to know what that means?
Sun's users are so daunted by the prospect of moving all their
applications to Solaris they are putting it off for as long as
they can. It is a complex move to an entirely new operating
system based on an incompatible version of UNIX (UNIX System V
Release 4), and, although Solaris has been around for some time
David Card, Director of Systems Research at IDC said he expects
users may wait to migrate until sometime in early 1994.
Sun announced in August that it was releasing the old Solaris 1.1C
on its newest workstation offerings. Customers were screaming
bloody murder about the effort required in migrating between the
two supposedly compatible versions of Sun's operating system that
Sun had to release the old Solaris 1.1 on the new workstations to
placate its angry installed base. Many customers will decide to
never migrate to Solaris 2.x because the transition would be
too difficult and too costly, even though they recognize they will
fall off Sun's technology curve. Adding to this dissatisfaction
with migration hassles is the fact that Sun charges customers more
for Solaris 2.x than they charged for the older Solaris 1.x.
Sun has an absolutely awful track record for protecting
customers' investments, and for migration to latest versions of
Solaris. It is a difficult, costly, painful migration that
requires major resource and time investment on the part of the
customer. DH Brown Associates' April 1993 report sums it up:
"..Solaris 2 represents a serious misstep. It provides only a
modest improvement on Solaris 1, but entails a painful migration
for users and developers. Had the changes provided significant
benefits, they might prove acceptable. This is not the case.
Sun's users are unlikely to feel comfortable with either the path
they must follow or what they will find when they arrive. Equally
perplexing, the difficult migration Solaris 2 entails may well
have been unnecessary. Many (users) will wait until Solaris 2.3
or 2.4, and many application packages become available - that is,
until 1994".
Sun may counter that now they have announced COSE compliance
arguments about operating systems standards are immaterial, this
is not the case. Although Sun have announced COSE compliance,
migrating Solaris to COSE will be a tremendous struggle for them -
Sun has no DCE or OpenView code, and although compliance with COSE
forces Sun to do the right thing strategically, it will be a
technical nightmare for them. Sun will be playing technical
catch-up for a long time to come, whilst HP and others move
further and further ahead with implementing COSE. Sun's only
apparent motivation to be COSE compliant is to allow them to
migrate their way out of a dead-end anyway.
vii) Sun's Poor Finances.
Sun's fiscal 1993 results have recently been published, and
although Sun's Q4 revenues pulled up what was a pretty poor set of
results, even their year-end figures and press announcement don't
paint a pretty picture for either shareholders or those concerned
with how much money is being invested in the company to keep Sun's
products current, and to put people in the field to support
customers' needs.
Annual revenues were up 20% from the previous year at $4.31
billion, but net income was down almost 10% to $1.49 per share.
(Until the end of Q3 the outlook was really depressing with
profits down 40% on fiscal 1992, and Sun's year-end press
announcement was full of news about the cutbacks they've been
forced to make - primarily cutbacks in people).
Sun claimed gross margin declines offset by tightly-controlled
growth as the primary reason for the better results, backing up
the view that Sun are cutting the number of people in the field.
Kevin Melia, Sun's Chief Financial Officer said in his 3rd quarter
statement:- "the gross margin percentage declined from the
prior-year quarter due to a mix of upgrade shipments, as well as
shipments of non-system components such as memory and storage. We
also experienced weakness in several European countries". That
sounds like they're only selling upgrades, RAM, and add-ons to
their installed base and are finding it difficult to get market
acceptance for new systems doesn't it?
Contrast this with HP's latest results - Q3 Fiscal '93 (ended July
31st) earnings rose 44% with revenue growing 23%. "Our profit
margins are up and earnings growing faster than revenue" said Lew
Platt. "We're pleased with this quarter's revenue growth. We did
a good job of ramping up shipments to respond to strong customer
demand for our new products".
In his half-year statement at the end of Q2, Robert Wayman, HP's
Chief Financial Officer said that operating expenses grew just 6%
which was a "superb outcome" considering HP has increased spending
to support strong gains in revenues and orders; and Lew Platt
reported that orders for HP's UNIX systems grew more than 40% with
HP being "on almost everyone's shortlist" of computer systems
suppliers. Our track record, market position, and current fiscal
results demonstrate a growing business and
faster-than-market-growth acceptance of HP's UNIX systems.
viii) SPARC's Future.
Although Sun designed the SPARC architecture, the implementation
and manufacture of the chip has been subcontracted to 3rd parties
like Cypress Semiconductor, Texas Instruments, Fujitsu(the largest
SPARC clone vendor), Toshiba, LSI, and NV Phillips. This shows
willing on Sun's part to open up the SPARC architecture to the
rest of the industry, but what's the reality? Sun has
consistently used product lead-time as an advantage to drive out
any clone maker who vies for the general purpose computer market.
At the outcry of its own SPARC International partners Sun promised
to change its behaviour to become more open, but IDC estimates
that even in 1995 85% of the demand for SPARC chips will come from
Sun. Sun's closed attitude to SPARC product availability makes a
joke of SPARC International and SPARC's wider use throughout the
industry.
Sun don't have their own chip fabrication facilities and therefore
use these 3rd parties to implement Sun's chip designs in silicon,
and to manufacture the chips. Sun used Fujitsu for the first
implementation, Cypress Semiconductor for the second, and now
Texas Instruments for the third. The result has been not only
poor time-to-market but also poor SPARC chip performance since
each new chip fabricator must learn about SPARC as they go along,
and there is no ongoing base of knowledge or building on previous
experience such as happens at HP with PA-RISC. HP completed the
entire PA7100 development and manufacture in 13 months - compared
to more than two years for Texas Instruments with SuperSPARC.
The delays suffered by TI in delivering SuperSPARC were due to
difficulties in design and fabrication, largely due to the
complexity of Sun's SuperSPARC chip design. Many industry
analysts/watchers believe it will be extremely difficult for Sun,
or their 3rd party fabricators, to drive SuperSPARC's clock speed
any higher than 50Mhz, thereby limiting the future potential of
the SPARC architecture. Cypress also makes a current SPARC chip
they've named HyperSPARC, it's an alternative to TI's SuperSPARC
and offers about the same performance, but Sun currently has no
plans to use it.
Why is all this important? Well, firstly the doubts over turning
up the clock speed on SuperSPARC should raise a big question in
your customers' minds - what do I do when I need to upgrade? And
secondly, who will make the chips that Sun are telling me I can
install in my Ss1000 or Sc2000 in future? There's no guarantee
these chips will be compatible with today's TI-made chips or the
systems they're used in - just look at Sun's past
backwards-compatibility with SunServer and 600MP systems, none of
which can use today's faster chips. Should your customers believe
what Sun is telling them?
Again, contrast this with HP's track record with PA-RISC, it's
upwards & downwards compatibility, it's industry-leading
performance, and HP's commitment to delivering leading-edge
technology year-on-year. The fact that HP owns its own
fabrication facilities and therefore controls the manufacturing of
its chips, that the design of future PA-RISC chips is carried out
in-house, and that HP has licensed PA-RISC to other companies for
wider market appeal; all adds up to a more stable, forseeable
future for customers choosing HP over Sun.
PRODUCT POSITIONING.
====================
__________________________________________________________________
When Sun bid... Position to win with Position to win
Performance/Expand- with Price/Perf-
ability. ormance.
__________________________________________________________________
SPARCserver 1000/1 Nova H30 Nova G30
SPARCserver 1000/2 Nova H50 Nova H40
SPARCserver 1000/4 Nova H60 Nova H60
SPARCserver 1000/6 Nova H70 Nova H70
SPARCserver 1000/8 Nova I70 Nova I70
__________________________________________________________________
Things to watch out for:-
i) As already discussed, coming from a workstation background Sun
have good compute performance, so be sure to include at least
one-third batch workload in any benchmark.
ii) Sun will try to get you to bid up to Emerald class systems.
Don't be tempted to do this - Sun have absolutely nothing to offer
that is in any way similar to Emerald, keep to the Nova family.
6. CONSULTANTS & INDUSTRY QUOTES.
===================================
"Sun has no experience in the data centre and niche vendors such
as Pyramid, Sequent, and Data General with their limited 3rd party
software offering, weak data centre management tools, and poor
corporate financials can only be considered for point,
non-strategic downsizing opportunities with low risk".
Workgroup Technologies Inc, December 1992.
"Sun has fallen off the price/performance curve, in part for
failing to keep SPARC up to the value offered by others."
Summit Strategies, Inc., August 1993
"Performance will be embarrassingly weak - as Sun found out when
it took exactly this approach with Solaris 1". Previous "four
processor configurations frequently achieved only poor performance
- almost NEVER approaching the 'linear performance increase' hoped
for, and sometimes even DECREASING total throughput compared to
the two processor configurations".
"We are skeptical, however, that Sun will achieve good relative
performance or linearity on its 2-20 processor SPARCcenter 2000
over the near term. The reason is simple: even vendors such as
DG, NCR, and SEQUENT with long histories of multiprocessing
experience have taken years to sufficiently tune their SMP UNIX
implementation to get near-linear performance increases as
processors are added. A belief that Sun can immediately achieve
what other vendors, despite their much greater experience, have
taken years to achieve is naive at the very least. Moreover,
systems comparable to the SPARCcenter 2000 such as those offered
by Sequent give notoriously uneven performance in practice."
DH Brown Associates, April 1993.
"Sun's past commercial conquests focussed on point solutions. Its
systems are not often deployed as enterprise systems and Sun does
not appear organised to sell to enterprise-level management".
Gartner Group, September 1992.
"Sun historically provided close to the WORST system management in
the UNIX world".
DH Brown Associates, April 1993.
"Solaris 2 sees great weaknesses in Security and Storage
Management. In Security, Solaris 2 offers only C1/UNIX security -
significantly lower than the C2 and B1 levels offered by most
competitors. Storage Management is an even bigger problem.
Essentially NO Storage Management tools such as disk managers or
enhanced backup/recovery systems are shipping for Solaris -
compared to the army of such crucial tools available for both
Solaris 1 and all competitive UNIX's".
DH Brown Associates, April 1993.
(HP-UX is)... "the most robust UNIX environment available"
Gartner Group,
"The high level of quality inherent in HP-UX".
DH Brown Associates, April 1993.
"Solaris 2's current weaknesses in multiprocessing implementation,
Storage Management, Security, and other areas illustrate the
company's macro challenge - its ongoing transition from a
technically focussed workstation company into a general systems
company that addresses both technical and commercial markets.
While begun, clearly much remains to be accomplished".
DH Brown Associates, April 1993.
"Large user sites have found that it takes weeks to port and debug
applications. Others anticipate an effort of several months".
Computerworld, June 1993.
"Sun views sales support, systems integration, and user needs
assessment as 3rd party, time consuming, costly activities that
divert from pressing needs to capture volume orders".
Gartner Group, September 1992.
"Most of ONC's services are technically weak in relation to those
of DCE. They are much less functional, less efficient, less
secure, harder to use, and harder to program. Though Sun has
begun to move beyond ONC with ONC+, DCE rather than ONC+ is THE
foundation for the next generation of high-function, heterogeneous
distributed computing".
DH Brown Associates, April 1993.
"Solaris 2 represents a serious misstep. It provides only a
modest improvement on Solaris 1, but entails a painful migration
for users and developers. Had the changes provided significant
benefits, they might prove acceptable. This is not the case.
Sun's users are unlikely to feel comfortable with either the path
they must follow or what they will find when they arrive. Equally
perplexing, the difficult migration Solaris 2 entails may well
have been unnecessary. Many (users) will wait until Solaris 2.3
or 2.4, and many application packages become available - that is,
until 1994".
DH Brown Associates, April 1993.
"The gross margin percentage declined from the prior-year quarter
due to a mix of upgrade shipments, as well as shipments of
non-system components such as memory and storage. We also
experienced weakness in several European countries".
Kevin Melia, Sun Microsystems, Q3 Fiscal Report, May 1993.
"The original SuperSPARC program was 12 to 18 months late by some
estimates. Worse yet, the frequency declined over time from as
high as 80 MHz in initial projections to 36 MHz in first
shipments."
Microprocessor Report, March 1993.
